Vtk master bump

[alliepiper]

Update code to work with new VTK master. See commit messages for details of changes.

This PR requires the following changes to the following repos:

• VTK needs to use Uvcdat master bump 27052015 6d673347 VTK#5
• uvcdat-testdata needs Update baselines for VTK bump. uvcdat-testdata#53

I'm missing one baseline from the testdata patch. I'm on arch, so the dv3d vector test fails for me (https://open.cdash.org/testDetails.php?test=339707560&build=3831990). Can someone run that test with this uvcdat/vtk and send me a baseline with the new text, but not the line. IIRC the bug doesn't affect ubuntu or mac.

This PR should also be manually tested, since the buildbots won't pick up the VTK change.

@doutriaux1 @chaosphere2112

[alliepiper]

@aashish24 The changes to vector plots (removing vcs2vtk.stripGrid) required a patch to VTK. The corresponding VTK merge request is here:

https://gitlab.kitware.com/vtk/vtk/merge_requests/242

[chaosphere2112]

@dlonie I've just finished a build with this update, so I'm going to run the tests + give you the new baseline.

[chaosphere2112]

@dlonie Here are my results... your updated baselines for test_vtk_ui_label_(top|x|y|left) are all incorrect. Don't know why the lines rendered that way on your system, but they should look like they used to (with a single pixel line).

[chaosphere2112]

@dlonie Looks like the dv3d tests are only different in the label, so they should be fine.

[chaosphere2112]

@dlonie Rest of the issues were all with this:

  File "/Users/samfries/bin/uvcdat/new_vtk/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/vcs/vcs2vtk.py", line 62, in putMaskOnVTKGrid
    vtk.vtkDataSetAttributes.GhostArrayName())
AttributeError: GhostArrayName

[chaosphere2112]

@dlonie And this:

  File "/Users/samfries/bin/uvcdat/new_vtk/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/vcs/vcs2vtk.py", line 79, in putMaskOnVTKGrid
    vtk.vtkDataSetAttributes.HIDDENPOINT
AttributeError: HIDDENPOINT

[chaosphere2112]

@dlonie Also has an issue with GhostArrayName on line 51 of vcs2vtk

[alliepiper]

@chaosphere2112

Re: Text alignment: I noticed the line was thicker on my system, but the text looks better aligned to my eye. Might be another platform issue? Feel free to send me some better ones if they look better on your system.

Re: vtkDataSetAttributes.* issues: Did you checkout the new VTK branch, and run make install in build/VTK-build? Those symbols are defined in the new VTK.

[doutriaux1]

@dlonie what's the status on this? Should I try and generate a new baseline? From Ubunut? From Mac?

[alliepiper]

I think Mac would be ideal, since I've run the tests on linux, and should check for platform specific stuff on Mac. Just be sure to build and install the VTK branch mentioned in the first post. Let me know if you run into any issues with it.

Thanks!

[doutriaux1]

works for me on mac now see:
https://open.cdash.org/buildSummary.php?buildid=3841120

[doutriaux1]

merging

[alliepiper]

What is the state of this at the moment? I see that this was merged, and then there's a closed, unmerged PR to revert it, and neither this branch nor the revert branch appear in master.

The VTK branch will have to be manually pushed, unless we have a new procedure for it. We've been just rebasing and pushing manually in the past to keep the modifications on top to more easily identify the difference with upstream.

[doutriaux1]

@dlonie I first merged this one then realized that would break everything since I couldn't update VTK. So I created a revert PR. But then I realize that @aashish24 would complain that the history is not clean. So I pushed back (-f) to the state before this PR goes in and closed the revert PR. Once you merged VTK we can use the re-issued PR to merge this ( #1357 ) which you merged so we should be good.